[elinks-users] Seg faults in elinks 0.11.3

Roy Millar awzg95 at dsl.pipex.com
Tue Apr 15 02:19:29 PDT 2008


Kalle Olavi Niemitalo writes :-

> There have been many bugs in the frames support, but I think this
> crash is not currently listed at bugzilla.elinks.cz.  Could you
> post a self-contained test case?

I'll try.
I'd need to know what was in the html being rendered, I suppose, but
the crash is almost immediate. Is the raw html kept in a buffer
somewhere?
Maybe tcpflow would show me, then it might be possible to guess at a
test case.
If I'm clever enough (doubtful; I'm long past my 'sell-before' date :-)

I could possibly edit the dump with a binary editor to overwrite any
traces of my security information with XXXXs, so that the bank
couldn't claim that I had breached their Terms and Conditions, then
email that to you/someone, if anyone would be kind enough to look at it.
If the actual dump might help?
Is there some secure way of emailing a large dump?

Thank you for the patch; I'll try to apply it (by hand if need be) to
0.11.3 and check the result.

To eliminate any possibility that the problem was caused by my use of
old libraries (though I'd guess that isn't likely), I tried to quickly
recompile elinks and js 1.7 using the LiveCD from Linuxfromscratch, but
unfortunately the link step fails for 'js' (not actually needed by
elinks, AFAIK).
The problem was an incompatibility between _xstat and __xstat, and
between __ctype-b, __ctype-b-toupper __ctype-b-lower and __ctype-b-loc

I'll try again, adding -Wl,_xstat=__xstat (etc.) to the link step to
see if that will kludge around the problem.

Or maybe just bypass the step that creates js.

Thank you again for your assistance.


-- 
R. T. Millar, awzg95 at dsl.pipex.com



More information about the elinks-users mailing list